VoxelBlockManager< Log2BlockWidth > Struct Template Reference

CPU counterpart of tools::cuda::VoxelBlockManager. Provides host-side decode of the inverse maps (sequential index -> leaf + voxel offset) for a single voxel block. The implementation is single-threaded per block and SIMD-accelerated (via util::shuffleDownMask and util::buildMaskPrefixSums); the caller is responsible for parallelism across blocks (e.g. OpenMP or nanovdb::util::forEach). More...

#include <nanovdb/tools/VoxelBlockManager.h>

Public Types
using	Base = VoxelBlockManagerBase<Log2BlockWidth>

Static Public Member Functions
template<class BuildT>
static util::enable_if< BuildTraits< BuildT >::is_index, void >::type	decodeInverseMaps (const NanoGrid< BuildT > *grid, const uint32_t firstLeafID, const uint64_t *jumpMap, const uint64_t blockFirstOffset, uint32_t *leafIndex, uint16_t *voxelOffset)
	Decode the inverse maps for a single voxel block on the host.

Static Public Attributes
static constexpr int	BlockWidth
static constexpr int	JumpMapLength
	number of uint64_t words per block in the jumpMap
static constexpr uint32_t	UnusedLeafIndex
	Sentinel written to leafIndex slots with no active voxel in this block.
static constexpr uint16_t	UnusedVoxelOffset
	Sentinel written to voxelOffset slots with no active voxel in this block.

Detailed Description

template<int Log2BlockWidth>
struct nanovdb::tools::VoxelBlockManager< Log2BlockWidth >

CPU counterpart of tools::cuda::VoxelBlockManager. Provides host-side decode of the inverse maps (sequential index -> leaf + voxel offset) for a single voxel block. The implementation is single-threaded per block and SIMD-accelerated (via util::shuffleDownMask and util::buildMaskPrefixSums); the caller is responsible for parallelism across blocks (e.g. OpenMP or nanovdb::util::forEach).

Member Typedef Documentation

Base

template<int Log2BlockWidth>

using Base = VoxelBlockManagerBase<Log2BlockWidth>

Member Function Documentation

decodeInverseMaps()

template<int Log2BlockWidth>

template<class BuildT>

util::enable_if< BuildTraits< BuildT >::is_index, void >::type decodeInverseMaps ( const NanoGrid< BuildT > * grid, const uint32_t firstLeafID, const uint64_t * jumpMap, const uint64_t blockFirstOffset, uint32_t * leafIndex, uint16_t * voxelOffset )

inlinestatic

Decode the inverse maps for a single voxel block on the host.

Given the VBM metadata for one block (firstLeafID and the block's slice of the jumpMap) and the block's base sequential offset, fills leafIndex[] and voxelOffset[] so that for each position p in [0, BlockWidth):

• leafIndex[p] = index of the leaf node containing sequential voxel (blockFirstOffset + p), or UnusedLeafIndex if that index is beyond the last active voxel.
• voxelOffset[p] = local (0..511) offset of that voxel within its leaf, or UnusedVoxelOffset.

The CPU analogue of the CUDA decodeInverseMaps. Single-threaded per block; SIMD is used internally. The caller is responsible for parallelism across blocks.

Template Parameters

BuildT

Build type of the grid (must be an index type)

Parameters

grid	Host-accessible OnIndex grid
firstLeafID	Index of the first leaf overlapping this block
jumpMap	Pointer to the JumpMapLength words for this block
blockFirstOffset	Sequential index of the first voxel in this block
leafIndex	Output array of length BlockWidth
voxelOffset	Output array of length BlockWidth

Member Data Documentation

BlockWidth

template<int Log2BlockWidth>

int BlockWidth

staticconstexpr

JumpMapLength

template<int Log2BlockWidth>

int JumpMapLength

staticconstexpr

number of uint64_t words per block in the jumpMap

UnusedLeafIndex

template<int Log2BlockWidth>

uint32_t UnusedLeafIndex

staticconstexpr

Sentinel written to leafIndex slots with no active voxel in this block.

UnusedVoxelOffset

template<int Log2BlockWidth>

uint16_t UnusedVoxelOffset

staticconstexpr

Sentinel written to voxelOffset slots with no active voxel in this block.